Construct a tokens object, either by importing a named list of characters from an external tokenizer, or by calling the internal quanteda tokenizer.

tokens() can also be applied to tokens class objects, which means that the removal rules can be applied post-tokenization, although it should be noted that it will not be possible to remove things that are not present. For instance, if the tokens object has already had punctuation removed, then tokens(x, remove_punct = TRUE) will have no additional effect.

tokens(
  x,
  what = "word",
  remove_punct = FALSE,
  remove_symbols = FALSE,
  remove_numbers = FALSE,
  remove_url = FALSE,
  remove_separators = TRUE,
  split_hyphens = FALSE,
  split_tags = FALSE,
  include_docvars = TRUE,
  padding = FALSE,
  concatenator = "_",
  verbose = quanteda_options("verbose"),
  ...,
  xptr = FALSE
)

Arguments

x

the input object to the tokens constructor; a tokens, corpus or character object to tokenize.

what

character; which tokenizer to use. The default what = "word" is the current version of the quanteda tokenizer, set by quanteda_options(okens_tokenizer_word). Legacy tokenizers (version < 2) are also supported, including the default what = "word1". See the Details and quanteda Tokenizers below.

remove_punct

logical; if TRUE remove all characters in the Unicode "Punctuation" [P] class, with exceptions for those used as prefixes for valid social media tags if preserve_tags = TRUE

remove_symbols

logical; if TRUE remove all characters in the Unicode "Symbol" [S] class

remove_numbers

logical; if TRUE remove tokens that consist only of numbers, but not words that start with digits, e.g. 2day

remove_url

logical; if TRUE removes URLs (http, https, ftp, sftp) and email addresses.

remove_separators

logical; if TRUE remove separators and separator characters (Unicode "Separator" [Z] and "Control" [C] categories)

split_hyphens

logical; if FALSE, do not split words that are connected by hyphenation and hyphenation-like characters in between words, e.g. "self-aware" becomes c("self", "-", "aware")

split_tags

logical; if FALSE, do not split social media tags defined in quanteda_options(). The default patterns are pattern_hashtag = "#\\w+#?" and pattern_username = "@[a-zA-Z0-9_]+".

include_docvars

if TRUE, pass docvars through to the tokens object. Does not apply when the input is a character data or a list of characters.

padding

if TRUE, leave an empty string where the removed tokens previously existed. This is useful if a positional match is needed between the pre- and post-selected tokens, for instance if a window of adjacency needs to be computed.

concatenator

character; the concatenation character that will connect the tokens making up a multi-token sequence.

verbose

if TRUE, print timing messages to the console

...

used to pass arguments among the functions

xptr

if TRUE, returns a tokens_xptr class object

Value

quanteda

tokens class object, by default a serialized list of integers corresponding to a vector of types.

Details

As of version 2, the choice of tokenizer is left more to the user, and tokens() is treated more as a constructor (from a named list) than a tokenizer. This allows users to use any other tokenizer that returns a named list, and to use this as an input to tokens(), with removal and splitting rules applied after this has been constructed (passed as arguments). These removal and splitting rules are conservative and will not remove or split anything, however, unless the user requests it.

You usually do not want to split hyphenated words or social media tags, but extra steps required to preserve such special tokens. If there are many random characters in your texts, you should split_hyphens = TRUE and split_tags = TRUE to avoid a slowdown in tokenization.

Using external tokenizers is best done by piping the output from these other tokenizers into the tokens() constructor, with additional removal and splitting options applied at the construction stage. These will only have an effect, however, if the tokens exist for which removal is specified at in the tokens() call. For instance, it is impossible to remove punctuation if the input list to tokens() already had its punctuation tokens removed at the external tokenization stage.

To construct a tokens object from a list with no additional processing, call as.tokens() instead of tokens().

Recommended tokenizers are those from the tokenizers package, which are generally faster than the default (built-in) tokenizer but always splits infix hyphens, or spacyr. The default tokenizer in quanteda is very smart, however, and if you do not have special requirements, it works extremely well for most languages as well as text from social media (including hashtags and usernames).

quanteda Tokenizers

The default word tokenizer what = "word" is updated in major version 4. It is even smarter than the v3 and v4 versions, with additional options for customization. See tokenize_word4() for full details.

The default tokenizer splits tokens using stri_split_boundaries(x, type = "word") but by default preserves infix hyphens (e.g. "self-funding"), URLs, and social media "tag" characters (#hashtags and @usernames), and email addresses. The rules defining a valid "tag" can be found at https://www.hashtags.org/featured/what-characters-can-a-hashtag-include/ for hashtags and at https://help.twitter.com/en/managing-your-account/twitter-username-rules for usernames.

For backward compatibility, the following older tokenizers are also supported through what:

"word1"

(legacy) implements similar behaviour to the version of what = "word" found in pre-version 2. (It preserves social media tags and infix hyphens, but splits URLs.) "word1" is also slower than "word2" and "word4". In "word1", the argument remove_twitter controlled whether social media tags were preserved or removed, even when remove_punct = TRUE. This argument is not longer functional in versions >= 2, but equivalent control can be had using the split_tags argument and selective tokens removals.

"word2", "word3"

(legacy) implements similar behaviour to the versions of "word" found in quanteda versions 3 and 4.

"fasterword"

(legacy) splits on whitespace and control characters, using stringi::stri_split_charclass(x, "[\\p{Z}\\p{C}]+")

"fastestword"

(legacy) splits on the space character, using stringi::stri_split_fixed(x, " ")

"character"

tokenization into individual characters

"sentence"

sentence segmenter based on stri_split_boundaries, but with additional rules to avoid splits on words like "Mr." that would otherwise incorrectly be detected as sentence boundaries. For better sentence tokenization, consider using spacyr.

Examples

txt <- c(doc1 = "A sentence, showing how tokens() works.",
         doc2 = "@quantedainit and #textanalysis https://example.com?p=123.",
         doc3 = "Self-documenting code??",
         doc4 = "£1,000,000 for 50¢ is gr8 4ever \U0001f600")
tokens(txt)
#> Tokens consisting of 4 documents.
#> doc1 :
#>  [1] "A"        "sentence" ","        "showing"  "how"      "tokens"  
#>  [7] "("        ")"        "works"    "."       
#> 
#> doc2 :
#> [1] "@quantedainit"              "and"                       
#> [3] "#textanalysis"              "https://example.com?p=123."
#> 
#> doc3 :
#> [1] "Self-documenting" "code"             "?"                "?"               
#> 
#> doc4 :
#> [1] "£"         "1,000,000" "for"       "50"        "¢"         "is"       
#> [7] "gr8"       "4ever"     "😀"       
#> 
tokens(txt, what = "word1")
#> Tokens consisting of 4 documents.
#> doc1 :
#>  [1] "A"        "sentence" ","        "showing"  "how"      "tokens"  
#>  [7] "("        ")"        "works"    "."       
#> 
#> doc2 :
#>  [1] "@quantedainit" "and"           "#textanalysis" "https"        
#>  [5] ":"             "/"             "/"             "example.com"  
#>  [9] "?"             "p"             "="             "123"          
#> [ ... and 1 more ]
#> 
#> doc3 :
#> [1] "Self-documenting" "code"             "?"                "?"               
#> 
#> doc4 :
#> [1] "£"         "1,000,000" "for"       "50"        "¢"         "is"       
#> [7] "gr8"       "4ever"     "😀"       
#> 

# removing punctuation marks but keeping tags and URLs
tokens(txt[1:2], remove_punct = TRUE)
#> Tokens consisting of 2 documents.
#> doc1 :
#> [1] "A"        "sentence" "showing"  "how"      "tokens"   "works"   
#> 
#> doc2 :
#> [1] "@quantedainit"              "and"                       
#> [3] "#textanalysis"              "https://example.com?p=123."
#> 

# splitting hyphenated words
tokens(txt[3])
#> Tokens consisting of 1 document.
#> doc3 :
#> [1] "Self-documenting" "code"             "?"                "?"               
#> 
tokens(txt[3], split_hyphens = TRUE)
#> Tokens consisting of 1 document.
#> doc3 :
#> [1] "Self"        "-"           "documenting" "code"        "?"          
#> [6] "?"          
#> 

# symbols and numbers
tokens(txt[4])
#> Tokens consisting of 1 document.
#> doc4 :
#> [1] "£"         "1,000,000" "for"       "50"        "¢"         "is"       
#> [7] "gr8"       "4ever"     "😀"       
#> 
tokens(txt[4], remove_numbers = TRUE)
#> Tokens consisting of 1 document.
#> doc4 :
#> [1] "£"     "for"   "¢"     "is"    "gr8"   "4ever" "😀"   
#> 
tokens(txt[4], remove_numbers = TRUE, remove_symbols = TRUE)
#> Tokens consisting of 1 document.
#> doc4 :
#> [1] "for"   "is"    "gr8"   "4ever"
#> 

if (FALSE) # using other tokenizers
tokens(tokenizers::tokenize_words(txt[4]), remove_symbols = TRUE)
tokenizers::tokenize_words(txt, lowercase = FALSE, strip_punct = FALSE) |>
    tokens(remove_symbols = TRUE)
#> Error in loadNamespace(x): there is no package called 'tokenizers'
tokenizers::tokenize_characters(txt[3], strip_non_alphanum = FALSE) |>
    tokens(remove_punct = TRUE)
#> Error in loadNamespace(x): there is no package called 'tokenizers'
tokenizers::tokenize_sentences(
    "The quick brown fox.  It jumped over the lazy dog.") |>
    tokens()
#> Error in loadNamespace(x): there is no package called 'tokenizers'